Interrogating overlap

[diazrenata]

What do I want to know about this metric?

[diazrenata]

WEIRD WEIRD WEIRD.

I tried computing the overlaps if you drop a mean somewhere in the body size range, sample all the individuals from the same normal distribution, and then assign them to species randomly with abundances according to the SAD.

And they're often pretty low!!!!

Just because you're sampling small numbers from a normal distribution.

So I don't know if this means...

• the overlap method isn't gonna work - everything is just going to sit on top of each other
• you need to weight it so that low-abundance species are downweighted more strongly than the harmonic mean achieves
• you need to force a larger bw for the kde

[diazrenata]

Update uses sum abundance weighting. There's probably a better way (a low abundance species should be really bad bc there's no knowing what its real kde is)

New update uses product abundance weighting, which amplifies the downweighting of pairings that include low abundance species/upweighting of high abundance pairings. (but not in the ecdf????)

I also coerced a version with completely even abundances, and as expected this had higher overlap values. It's an issue with low sampling, low certainty, and extremely pointy kdes for the low-abundance species.